Re: Kernel quirks on QNAP TS-109 (Marvell orion)

[Arnaud Patard (Rtp) <arnaud.patard@rtp-net.org>]

Martin Michlmayr <tbm@cyrius.com> writes:

> * Matthieu CERDA <kegeruneku+debianarm@ironflake.org> [2019-07-26 00:17]:
>>   * On stretch stock kernel (4.9.0-9-marvell), qcontrol does not work:
>>     calling it with "qcontrol --direct buzzer" outputs no error, but
>>     does nothing, and the status led stays red/green after system has
>>     booted.
>
> There are different potential causes for this but I think I've seen
> this before myself and this particular issue hasn't been reported.
>
>>   * On stretch backports kernel (4.19.0-0.bpo.5-marvell), qcontrol does
>>     not work as well, and more annoyingly, no Ethernet interface gets
>>     initialized. dmesg shows a kernel oops. (copy attached)
>
> This sounds like https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=908712

As noted in the bug, might worth reporting to Andrew. iirc ts109 are
devices without device tree and I'm curious to see how the kernel handle
things like of_clk_get(pdev->dev.of_node, i) on a system without DT.
of_clk_get() is coded like this:

struct clk *of_clk_get(struct device_node *np, int index)
{
        return __of_clk_get(np, index, np->full_name, NULL);
}

afair in systems without DT, of_node is null, np->full_name has high
chances to crash. Again, as noted by the bug, it's likely coming from
96cb4342382290c935d933a08feb57d6d0183071 which is replacing
the devm_clk_get() call to of_clk_get().

Might worth trying to build a kernel with code looking like
(not even compile tested):

if (pdev->dev.of_node) {
        for (i = 0; i < ARRAY_SIZE(dev->clk); i++) {
            dev->clk[i] = of_clk_get(pdev->dev.of_node, i);
...
        }
        if (!IS_ERR(of_clk_get(pdev->dev.of_node, ARRAY_SIZE(dev->clk))))
        ...
        
else {
     dev->clk[0] = clk_get(&pdev->dev, NULL);
     if (!IS_ERR(dev->clk[0]))
        clk_prepare_enable(dev->clk[0]);
}


Arnaud.